Although antidepressant medications are effective in reducing depressive symptoms among adults, concerns recently have re-emerged that these medications may increase suicide risks during times of treatment initiation or dosage change. The FDA has issued a "black box" warning for antidepressants when used by children and adolescents and advised close monitoring of patients of all ages when beginning or changing doses. Current FDA warnings are based on suicidal behaviors rather than completed suicide. However, these intermediate outcomes are often difficult to categorize and may not be on a complete continuum with completed suicide. Outcomes of treatment in the community may also differ from outcomes in controlled trials in which the use of adjunctive medications are sharply limited. We will use the VA National Registry for Depression (NARDEP), Medicare, and National Death Index data to examine the relationships between antidepressants, concomitant treatments, and completed suicide. NARDEP includes longitudinal data for over 1,500,000 patients diagnosed with depression from fiscal year (FY) 1997 forward, including data on demographics, health services use (from FY97 forward), and pharmacy use (from FY99 forward). Approximately 994,000 individuals will meet inclusion criteria for the proposed study, with over 2,800,000 person-years of data. We will evaluate potential differences in rates of suicide and suicide attempts among patients receiving: a) treatment with different antidepressant medication groups, b) treatment with four most commonly prescribed agents within the SSRI class and the three most commonly prescribed alternatives to SSRI agents, c) adjunctive agents used with antidepressants to augment treatment response, and d) supplemental hypnotic agents used with antidepressants to address insomnia. Study analyses will first use Cox-proportional hazards models with time-varying treatment exposures and covariates. We will also use Cox models with inverse probability of treatment weighting (IPTW) and instrumental variables to address potential treatment selection biases. In addition to the primary cohort analyses, we will conduct two nested case-control studies, using medical record data, to conduct a detailed assessment of potential selection biases that are not observable in the administrative data. These sub- studies will inform the cohort analyses and allow us to appropriately interpret study results. Clinicians, policymakers, and patients urgently need additional information about relative effectiveness of alternative depression treatments in reducing suicide. We anticipate that the proposed project will advance knowledge in this critical area and inform future efforts to use organizational datasets to examine sentinel events, such as suicide. [unreadable] [unreadable] [unreadable]